[PATCH v3] perf tools: Support -x for perf stat report
From: Andi Kleen
Date: Sun Nov 01 2020 - 01:36:56 EST
Add support for the -x, option to enable CSV output with perf stat
report. Useful to parse the information with other programs.
% perf stat record --quiet -a -I 1000 sleep 5
% perf stat report -x,
1.000838672,4003.55,msec,cpu-clock,4003548736,100.00,,
1.000838672,11243,,context-switches,4003631885,100.00,0.003,M/sec
1.000838672,1682,,cpu-migrations,4003672150,100.00,0.420,K/sec
1.000838672,13244,,page-faults,4003697471,100.00,0.003,M/sec
1.000838672,2953214077,,cycles,4003715495,100.00,0.738,GHz
1.000838672,4380820799,,instructions,4003738270,100.00,1.48,insn per cycle
1.000838672,809865653,,branches,4003760616,100.00,202.287,M/sec
1.000838672,12439843,,branch-misses,4003780785,100.00,1.54,of all branches
...
Signed-off-by: Andi Kleen <ak@xxxxxxxxxxxxxxx>
---
v2: Fix default output (Jiri). Also handle \t as special value like the
original parser (Andi)
v3: Use DEFAULT_SEPARATOR
---
tools/perf/builtin-stat.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 743fe47e7a88..9fcc7351ce43 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1988,6 +1988,8 @@ static int __cmd_report(int argc, const char **argv)
"aggregate counts per numa node", AGGR_NODE),
OPT_SET_UINT('A', "no-aggr", &perf_stat.aggr_mode,
"disable CPU count aggregation", AGGR_NONE),
+ OPT_STRING('x', "field-separator", &stat_config.csv_sep, "separator",
+ "print counts with custom separator"),
OPT_END()
};
struct stat st;
@@ -2002,6 +2004,13 @@ static int __cmd_report(int argc, const char **argv)
input_name = "perf.data";
}
+ if (strcmp(stat_config.csv_sep, DEFAULT_SEPARATOR)) {
+ stat_config.csv_output = true;
+ if (!strcmp(stat_config.csv_sep, "\\t"))
+ stat_config.csv_sep = "\t";
+ stat_config.big_num = false;
+ }
+
perf_stat.data.path = input_name;
perf_stat.data.mode = PERF_DATA_MODE_READ;
--
2.28.0